Integer Discrete Flows and Lossless Compression

Authors: Emiel Hoogeboom, Jorn Peters, Rianne van den Berg, Max Welling

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments show that IDFs are competitive with other flow-based generative models. Furthermore, we demonstrate that IDF based compression achieves state-of-the-art lossless compression rates on CIFAR10, Image Net32, and Image Net64.
Researcher Affiliation Collaboration Emiel Hoogeboom Uv A-Bosch Delta Lab University of Amsterdam Netherlands e.hoogeboom@uva.nl Jorn W.T. Peters Uv A-Bosch Delta Lab University of Amsterdam Netherlands j.w.t.peters@uva.nl Rianne van den Berg University of Amsterdam Netherlands riannevdberg@gmail.com Max Welling Uv A-Bosch Delta Lab University of Amsterdam Netherlands m.welling@uva.nl
Pseudocode No The paper includes a diagram (Figure 2) but no structured pseudocode or algorithm blocks.
Open Source Code Yes Code to reproduce the experiments is available at https://github.com/jornpeters/integer_discrete_flows.
Open Datasets Yes To test the compression performance of IDFs, we compare with a number of established lossless compression methods... on CIFAR10, Image Net32 and Image Net64. In particular, the ER + BCa histology dataset [18] is used, which contains 141 regions of interest scanned at 40 .
Dataset Splits No The paper mentions training on 'train data' and reporting on 'test data'. It also states 'For the exact treatment of datasets and optimization procedures, see Section D.4', but does not explicitly specify validation dataset splits within the provided text.
Hardware Specification No No specific hardware details (e.g., exact GPU/CPU models or processor types) used for running experiments are provided in the paper.
Software Dependencies No The paper mentions using 'Py Torch' [28] but does not provide specific version numbers for any software dependencies.
Experiment Setup No The paper states, 'The specific architecture details for each experiment are presented in Appendix D.1.' and 'For the exact treatment of datasets and optimization procedures, see Section D.4.', indicating setup details are deferred to appendices not provided in the main text.